Mushroom-rl

- Added missing resources to setup
- Moved humanoid meshes into meshes folder (was Geometry)
- Improved makefile in order to ensure proper build before uploading

Fixed source distribution by adding a missing .pyx file.

- Implemented PEP 517-518;
- Improved saving system:
* Added Serializable interface that can be used by any Mushroom class;
* Approximator is now serializable;
* Fixed issue with save/load attributes;
* Everything is saved in a zip file with .msh extension.
- Implemented cmac regressor;
- Step callback added in core;
- callbacks_episode renamed in callbacks_fit;
- Distributions are now serializable;
- Added computation of entropy for distributions;
- Policies are now serializable;
- Added BoltzmannTorchPolicy;
- Added LQR solver;
- Added PyBullet environment;
- Fixed issue in Gym environment when PyBullet import was missing;
- LQR bug fixed and generate function improved;
- Improved Mujoco environment;
- Cythonization and cleanup of HumanoidGait environment;
- TRPO bug fixed;
- Added RandomFourierBasis;
- Fixed pendulum_trust_region example.

- Save and load of the agent;
- Online data and performance visualization using pyqtgraph;
- Normalization function added as preprocessors;
- Step callback added;
- Humanoid gait environment.

MushroomRL is the new name of the library. Code, doc, and config files have been updated accordingly.
The package is now available on PyPI and can be installed using pip.

- Improved test suit;
- Implemented Deep RL algorithms;
- Improved visualization;
- Improved code quality;
- Several bug fixes;
- Refactoring of Torch approximator.